Tese de Doutoramento Medicina 2024 Faculdade de Ciências Médicas, Universidade NOVA de Lisboa

Abstract Background: While the prognosis of patients with idiopathic pulmonary fibrosis (IPF) is well established, the prognosis for other interstitial lung diseases (ILDs) is highly variable. Despite immunosuppressive treatment, approximately 30% of non-IPF ILD patients will experience disease progression. While lung function decline and increased fibrosis on computed tomography (CT) scan are known determinants of mortality, there are currently no reliable methods to predict, in a timely manner, which of these patients will progress, and which will remain stable. Given the emergence of new therapeutic options, there is an urgent need for non-invasive techniques that can effectively predict prognosis and guide treatment decisions, particularly for patients with non-IPF ILD. Although magnetic resonance imaging (MRI) has demonstrated to be effective in differentiating inflammation from fibrosis in other organs, its application in the lungs and particularly in ILD is still in the early stages. There is a lack of consensus on how to evaluate MRI signal intensity values for ILD characterization, with some existing methods being prone to significant variability among observers. Furthermore, few studies have explored the prognostic potential of MRI in ILD. Objectives: The primary aim of this research was to assess the utility of MRI as an additional imaging method for staging disease severity and predicting prognosis in patients with non-IPF ILDs. The specific objectives of this study were as follows: Task 1: To assess the quality of MR images and determine their accuracy in the identification of ILD features in comparison with CT; Task 2: To determine the most reproducible method to analyze ILD with MRI; Task 3: To correlate MRI features with CT and functional parameters at baseline; Task 4: To correlate MRI features with ILD progression and the evolution of lung function tests (LFTs) at 12 months. Methods: In this prospective observational cohort study (NMS no60/2018/CEFCM; CHULC 618/2018; HL CES/20/2020/ME), patients with non-IPF ILDs underwent baseline MRI evaluations, along with standard clinical examinations, CT and lung function tests (LFTs). Patients were then followed by their referring physicians for a total of 12 months, with regular follow-up visits occurring at least every 6 months. CT scans and LFTs were performed at study enrolment and again after a 12-month follow-up. T2-weighted images (T2-WI) were acquired by axial free-breathing respiratory-gated fat-suppressed BLADE sequence, a PROPELLER (periodically rotated overlapping parallel lines with enhanced reconstruction)-equivalent of Siemens Healthineers, and T1-weighted images (T1-WI) by coronal end-expiratory breath-hold fat-suppressed VIBE (volumetric interpolated breath-hold examination) sequences, before and at time points T1, T3, T5 and T10 minutes after gadolinium administration. For Task 1, T2-W and T1-W dynamic study images were evaluated for the presence of artifacts; signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR) and noise were also calculated. The identification of ground-glass opacities (GGO)/reticulation, traction bronchiectasis, cystic changes and consolidation was compared between MRI and CT, using CT as the gold-standard. Regarding Task 2, MRI images of 18 patients were evaluated with two different methods: a qualitative evaluation [visual assessment and measurement of few regions of interest (ROIs); evaluations were performed independently by 5 radiologists and three times by 1 radiologist] and a quantitative segmentation-based analysis with software extraction of signal intensity values (evaluations were performed independently by 2 radiologists and twice by 1 radiologist). ILD was classified as inflammatory or fibrotic, based on previously described imaging criteria, for comparative purposes. After this preliminary evaluation, the segmentation-based quantitative evaluation was selected for the analysis of the complete cohort. As for Task 3, manual segmentation with delineation of ILD affected regions was performed for T2-WI and for all acquisitions in the T1-WI dynamic studies. Segmentation of the total lung volume was also performed for T2-WI and small ROIs were placed in the paraspinal muscles and normal lung on both sequences. Signal intensity values were extracted by a dedicated software. For T2-WI, the following parameters were calculated: the extent of ILD, determined as a percentage of the total lung volume (%ILD volume); ratios between the signal intensity of ILD and that of the normal lung (SIILD/SInormal lung) and of the paraspinal muscles (SIILD/SImuscle); SIILD/SInormal lung and SIILD/SImuscle ratios were multiplied by %ILD volume to generate signal-to-volume scores (muscle and normal lung, respectively). Percentage of signal intensity (%SI) at each time-point, time to peak enhancement, and percent relative enhancement of ILD in comparison with normal lung (%SIILD/normal lung) were calculated for T1-WI, among other parameters. MRI features were correlated with LFTs and CT parameters at baseline. Concerning Task 4, baseline MRI parameters were correlated with the diagnosis of disease progression at 12 months (based on clinical findings, LFTs and follow-up CT and with the variation in percent predicted forced vital capacity (%FVC) and percent predicted diffusing capacity for carbon monoxide (%DLCO) obtained after a 12- months interval. Results: 25 patients performed MRI evaluation at baseline (21 with complete T2-WI and T1-WI study). The mean age was 62.6 years, 68% were females. Mean disease duration was 2.6 ± 2.97 years. 76% of patients had a diagnosis of non-specific interstitial pneumonia (NSIP) [(58% associated with connective tissue diseases (CTD)], and a minority of patients had a diagnosis of fibrotic hypersensitivity pneumonitis (8%), CTD variant fibrosis (8%) and unclassifiable ILD (8%). Task 1: T2-W images (n=25) were classified as having mostly low-grade artifacts (64%). On T1-WI (n=21), the images of one patient were excluded due to high-grade artifacts; the remaining T1-WI studies showed mostly low-grade artifacts (57.1%). GGO/reticulation were correctly identified in all MRI studies (100% accuracy). Traction bronchiectasis were seen on MRI in 12 patients and cystic changes in 6 (both with 9 false negatives, 64% accuracy). Task 2: Regarding the qualitative evaluation, intra-observer agreement was excellent (κ= 0.92, p<0.05) for T2-WI and fair (κ= 0.29, p<0.05) for T1 dynamic study, while inter observer agreement was moderate (κ= 0.56, p<0.05) and poor (κ= 0.11, p=0.18), respectively. In contrast, upon qualitative segmentation-based analysis, intra- and inter-observer agreement were excellent for T2-WI (κ=0.886, p<0.001; κ=1.00, p<0.001; respectively); for T1-WI, intra-observer agreement was excellent (κ= 0.87, p<0,05) and inter-observer agreement was good (κ= 0.75, p<0.05). Task 3: A significant positive correlation was seen between T2-WI %ILD volume and CT scores (r=0.62-0.87, p≤0.001). Regarding LFTs, a significant negative correlation was seen between T2-WI %ILD volume with both %FVC and %DLCO at baseline (r=-0.44 to -0.67, p<0.05). As for T1-WI parameters, only time to peak enhancement showed a significant positive correlation with %FVC (r=0.50, p<0.05). Task 4: Three of the 24 patients that completed follow-up fulfilled criteria for disease progression. Baseline T2-WI SIILD/SInormal lung was higher for the progression group (p=0.052). T2-WI SIILD/SInormal lung and T1-WI %SIILD/normal lung at T1 were positively correlated with decline in %FVC (r=0.495, p=0.014 and r=0.489, p=0.034, respectively) at 12- months. Both MRI parameters were independent predictors of %FVC decline. Conclusions: By employing conventional MRI sequences and a segmentation-based quantitative analysis, we were able to identify parameters that correlate with %FVC decline at 12 months, specifically SIILD/SInormal lung ratio and relative enhancement of ILD in comparison with normal lung in the first minute. Baseline T2-WI SIILD/SInormal lung ratio was also higher in patients with diagnosis of progression, although with marginal significance, requiring further confirmation through larger-scale studies In relation to each specific task: Task 1: Good quality T2-W and T1-W MRI images were obtained, characterized by adequate contrast-to-noise and signal-to-noise ratios. Fat-saturated T2-WI BLADE sequence allowed for precise identification of GGO/reticulation. However, the accuracy for identification of traction bronchiectasis and cystic changes was relatively diminished. Task 2: The quantitative segmentation-based MRI analysis was more reproducible than the qualitative evaluation with visual assessment and measurement of few ROIs. Task 3: T2-WI %ILD volume was found to align with the overall extent of ILD observed on CT scans. Moreover, T2-WI %ILD volume exhibited a negative correlation with baseline %FVC and %DLCO values. Task 4: Patients who experienced disease progression showed higher baseline T2-WI signal intensity compared to that of normal lung (with marginal significance). Additionally, both this ratio and the T1-WI relative enhancement of ILD in comparison to normal lung during the first minute were identified as significant independent predictors of %FVC decline at 12 months